1
Levonorgestrel-releasing intrauterine system use is associated with a
1
decreased risk of ovarian and endometrial cancer, without increased
2
risk of breast cancer. Results from the NOWAC Study
3 4
Mie Jareid1*, Jean-Christophe Thalabard2, Morten Aarflot1, Hege M. Bøvelstad1, Eiliv Lund1, Tonje 5
Braaten1 6
1) Department of Community Medicine, Faculty of Health Sciences, UiT – The Arctic University of 7
Norway, Tromsø, Norway 8
2) Applied Mathematics Lab, UMR CNRS 8145, Paris Descartes University, USPC 9
E-mail addresses:
10
MJ: mie.jareid@uit.no 11
JCT: Jean-Christophe.Thalabard@parisdescartes.fr 12
MAA: morten.arflot@uit.no 13
HMB: hege.m.bovelstad@uit.no 14
EL: eiliv.lund@uit.no 15
TB: tonje.braaten@uit.no 16
17
*Corresponding author: Mie Jareid, Department of Community Medicine, Faculty of Health Sciences, 18
UiT The Arctic University of Norway, Pb 6070 Langnes, NO-9037 Tromsø, Norway; Telephone: +47 19
77625209; E-mail: mie.jareid@uit.no 20
Disclaimer: Some of the data in this article are from the Cancer Registry of Norway. The Cancer Registry 21
of Norway is not responsible for the analysis or interpretation of the data presented.
22
Word count: 4044, Abstract: 250 23
Keywords: Epidemiology; prospective cohort study; ovarian neoplasms; uterine neoplasms; breast 24
neoplasms; intrauterine device; intrauterine device, medicated; levonorgestrel; levonorgestrel-releasing 25
intrauterine system; hormonal contraceptives 26
Abbreviations used: BMI, body mass index; CI, confidence interval; ICD, International Classification of 27
Diseases; LNG-IUS, levonorgestrel-releasing intrauterine system; NOWAC, Norwegian Women and 28
Cancer; OC, oral contraceptives; PY, person-years; RR, relative risk; SD, standard deviation; SIR, 29
standardized incidence ratio 30
31
2 ABSTRACT
32
Objective Women with ovarian cancer have poor survival rates, which have proven difficult to improve;
33
therefore primary prevention is important. The levonorgestrel-releasing intrauterine system (LNG-IUS) 34
prevents endometrial cancer, and recent studies suggested that it may also prevent ovarian cancer, but 35
with a concurrent increased risk of breast cancer. We compared adjusted risks of ovarian, endometrial, 36
and breast cancer in ever users and never users of LNG-IUS.
37
Methods Our study cohort consisted of 104 318 women from the Norwegian Women and Cancer Study, 38
9144 of whom were ever users and 95 174 of whom were never users of LNG-IUS. Exposure information 39
was taken from self-administered questionnaires, and cancer cases were identified through linkage to 40
the Cancer Registry of Norway. Relative risks (RRs) with 95% confidence intervals (CIs) were estimated 41
with Poisson regression using robust error estimates(1).
42
Results Median age at inclusion was 52 years and mean follow-up time was 12.5 (standard deviation 43
3.7) years, for a total of 1 305 435 person-years. Among ever users of LNG-IUS there were 18 cases of 44
epithelial ovarian cancer, 15 cases of endometrial cancer, and 297 cases of breast cancer. When ever 45
users were compared to never users of LNG-IUS, the multivariable RR of ovarian, endometrial, and 46
breast cancer was 0.53 (95% CI: 0.32, 0.88), 0.22 (0.13, 0.40), and 1.03 (0.91, 1.17), respectively.
47
Conclusion In this population-based prospective cohort study, ever users of LNG-IUS had a strongly 48
reduced risk of ovarian and endometrial cancer compared to never users, with no increased risk of 49
breast cancer.
50 51
3 52
INTRODUCTION 53
In 2012, ovarian cancer caused an estimated 152 000 deaths worldwide (2). The cumulative risk of 54
ovarian cancer until age 75 is 1.3% in Norway and is similar in the United States (3, 4). The symptoms of 55
ovarian cancer are vague, and there is no screening test. This has led to problems of late diagnosis and a 56
5-year survival of less than 50% (5). Thus, ovarian cancer ranks eighth in cancer incidence, but fifth in 57
cancer mortality among women (4). Primary prevention therefore remains the best available measure 58
against ovarian cancer (5).
59
Risk of ovarian cancer is reduced by 15-29% for every 5 years of oral contraceptive (OC) use, and 60
globally, OC use prevents an estimated 30 000 cases of ovarian cancer each year (6). Long-term OC use 61
also reduces the risk of endometrial cancer, with 5-9 years of use reducing the risk by 34% (7). However, 62
OC use increases the risk of breast cancer up to 38% with more than 10 years use, and for minimum 5 63
years after cessation (8, 9) in addition to carrying other health risks. Prescribing OCs for ovarian cancer 64
prevention to women who do not need contraception is not recommended (10).
65
The levonorgestrel-releasing intrauterine system (LNG-IUS) was introduced in Norway in 1994. In 66
the Nordic countries, LNG-IUS is the second-most used form of contraception after OCs, and it is the 67
most commonly used form of long-acting reversible contraception (11). Recently, three Finnish studies 68
have shown that, compared to the general population, LNG-IUS users have a standardized incidence 69
ratio (SIR) of 0.59 for ovarian cancer and 0.46 for endometrial cancer (12, 13), but also an increased risk 70
of ductal and lobular breast cancer (SIR 1.20 and 1.33 respectively, increasing to SIR 1.37 and 1.73 with 71
more than 5 years of use) (14). However, these studies did not adjust for other hormonal risk factors.
72
Our study aim was to combine self-reported information on OC use and reproductive factors from 73
the Norwegian Women and Cancer (NOWAC) Study, with registry-based follow-up of cancer cases to 74
compare adjusted risks of ovarian, endometrial, and breast cancer in ever users and never users of LNG- 75
4 IUS. We also included estimates of the reduction in the risk of endometrial cancer in this nationally 76
representative cohort, given the well-known preventive effect of LNG-IUS use on this cancer (15).
77
METHODS 78
Study cohort 79
The NOWAC Study is a population-based prospective cohort study designed to investigate the 80
association between hormone use and hormone-dependent female cancers (16). During 1991-2007, 81
women born between 1927 and 1965 were randomly selected from the Norwegian Population Registry 82
and were sent a questionnaire along with a letter that explained the study. Those who returned a 83
completed questionnaire were enrolled. Statistics Norway replaced participants’ names and personal 84
identification numbers with serial numbers for use by researchers. Recruitment took place in two waves:
85
102 540 participants were enrolled in 1991-1997 (response rate 57%), and 63 232 participants in 2003- 86
2006 (response rate 48.4%). The external validity of the NOWAC Study was found to be good (17).
87
Follow-up information has been collected up to two times after enrollment.
88
The NOWAC questionnaires targeted LNG-IUS use as from 1998 by the question: Have you ever used 89
a hormone intrauterine device? A total of 145 320 women completed a questionnaire during 1998-2006, 90
either at enrollment or as part of follow-up. From these, we excluded 33 182 that either did not answer 91
the question on hormone intrauterine device or had a hysterectomy or oophorectomy; 4813 that either 92
had prevalent cancer or died or emigrated before the start of follow-up; 2938 that indicated LNG-IUS 93
use before the device was available in Norway, and seven for technical reasons. Thus the final study 94
cohort consisted of 104 380 women, of which 9146 were ever users of LNG-IUS.
95
Exposure assessment 96
In addition to questions on LNG-IUS (ever use, duration of use, age at first use, current use), we 97
identified eight exposure variables associated with ovarian, endometrial, or breast cancer (18), 98
regardless of their association with LNG-IUS use: age at start of follow-up (41-76 years, in 4-year 99
5 increments), body mass index at enrollment (BMI, <25 kg/m2, ≥25 kg/m2), physical activity level at 100
enrollment (very low, low, intermediate, high, very high), maternal history of breast cancer (yes, no), 101
age at menarche (<12, 12-14, ≥15), ever use of OCs (yes, no), parity (0, 1-2, 3-4, ≥5), and menopausal 102
status at start of follow-up (pre, peri, post, unknown). Unknown menopausal status was given to those 103
who used hormone replacement therapy, those who indicated that menses had stopped due to 104
“medication, illness, exercise, or other” and to those who did not answer the question.
105
Outcomes 106
Primary cancers were identified through linkage to the Cancer Registry of Norway using the 107
International Classification of Diseases, Revision 7 (ICD-7) codes. All citizens were identified by their 108
personal identification number upon contact with health care providers, who are obliged to report all 109
cancer cases to the Cancer Registry of Norway. Cases were defined as cancer of the ovary including the 110
fallopian tube (ICD-7 code 175), cancer of the uterine corpus (ICD-7 code 172), and cancer of the breast 111
(ICD-7 code 170). In order to restrict the analyses to epithelial ovarian cancer and endometrial cancer, 112
non-carcinoma cancers of the ovary and uterine corpus were excluded from the analyses (n=62). Deaths 113
and emigrations were identified through the Cause of Death Registry and Statistics Norway. Follow-up 114
ended on 31 December 2015.
115
Statistical analysis 116
We calculated person-years (PY) of follow-up from the date of entrance into, until the date of exit from 117
the study. Exit date was defined as the date of cancer diagnosis, emigration from Norway, death, or end 118
of follow-up, whichever occurred first. We used chi-squared tests of independence to compare the 119
characteristics of ever users and never users of LNG-IUS, and to compare selected characteristics of 120
responders and non-responders of the question on LNG-IUS use.
121
We calculated crude cancer incidence rates with 95% confidence intervals (CIs) assuming a Poisson 122
distribution. Relative risks (RRs) and their 95% CIs were estimated with Poisson regression using a robust 123
6 error estimate. Adjusted RR models were built in a stepwise backward manner by removing
124
nonsignificant covariates from the full model, with listwise deletion of participants with missing 125
information. Model fit was assessed by testing the deviance versus its assumed chi-squared distribution.
126
Statistical significance was defined as a test resulting in a p-value <0.05. We performed an additional 127
analysis of the association between LNG-IUS use and endometrial cancer, stratified by ever OC use (yes, 128
no), and did a Wald test of heterogeneity between the resulting RRs. We performed two additional 129
analyses of the association between LNG-IUS use and breast cancer: stratified by duration of use (≤5 and 130
>5 years), and stratified into current and former users at the start of follow-up.
131
The analyses were performed in SAS software version 9.4 (SAS Institute, Inc., Cary, NC, USA ).
132
Ethics 133
The Regional Ethics Committee, REK Nord, approved the NOWAC Study. Written information was 134
provided to the participants, and return of a completed questionnaire was considered as consent to 135
participate. Data storage is in compliance with the rules of the Norwegian Data Inspectorate.
136 137
7 RESULTS
138
Median age at inclusion was 52 years. Mean follow-up time was 12.5 (standard deviation [SD] 3.7) years 139
for a total of 1 305 435 PY. Among all ovarian and uterine corpus cancers, respectively 4% and 5% were 140
non-carcinoma cancers and were excluded. Of the women in the study cohort, 9144 (9%) reported LNG- 141
IUS use during or prior to the data collection period (1998-2007). Among ever users of the LNG-IUS, 85%
142
reported the duration of use. Median age at starting LNG-IUS use was 44 years, and median duration 143
was 4 years, with 59% having used LNG-IUS for between 2 and 6 years. Compared to never users, ever 144
users of LNG-IUS were younger at start of follow-up (Table 1).
145
The percentage of women that reported high or very high physical activity level was slightly higher 146
among ever users of LNG-IUS (38% versus 30% of never users) (Table 1). Ever use of OCs was more 147
common among ever users of LNG-IUS (71%) than never users (55%), and nulliparity was more common 148
among never users of LNG-IUS (10% vs. 3% among ever users). Menopausal status at start of follow-up 149
was significantly different between the groups of LNG-IUS use, with 60% of never users reporting that 150
they were postmenopausal, compared to 33% of ever users. Thirty percent (n=2753) of ever users had 151
unknown menopausal status, and of these, 85% were using LNG-IUS at the start of follow-up.
152
Non-responders to the LNG-IUS question (n=15442) differed significantly from the study cohort on 153
all variables checked. Most notably they had a lower proportion of nullipara (Supplementary table S1).
154 155
Levonorgestrel-releasing intrauterine system and cancer incidence 156
Table 2 displays cancer incidences and risk estimates. The crude incidence rate of ovarian cancer among 157
never users of LNG-IUS was 38.1 (95% CI: 34.7, 41.8). The crude incidence rate of ovarian cancer among 158
ever users of LNG-IUS was 16.7 per 100 000 PY (95% CI: 9.9, 26.4), with an age-adjusted RR of 0.49 (95%
159
CI: 0.30, 0.82) for ever versus never users. The final model for ovarian cancer included three significant 160
covariates: age at start of follow-up, ever use of OCs, and menopausal status at start of follow-up. Parity 161
8 was not significant in the model building, but qualified as a confounder and was included in the model.
162
Adjustment for these covariates hardly changed the risk estimates (multivariable-adjusted RR 0.53 (95%
163
CI: 0.32, 0.88)).
164
The reported duration of LNG-IUS use varied from less than 1 year to 14 years, with the latter value 165
corresponding to the time difference between the introduction of the LNG-IUS in 1994 in Norway and 166
the date of the last questionnaire (2008). There were 18 cases of ovarian cancer among ever users of 167
LNG-IUS; 14 of these cases occurred in women who had been using LNG-IUS for less than 7 years, and 3 168
in women who did not report duration of use. Due to the low number of cases, duration-response 169
estimates were not calculated.
170
The largest risk reduction was observed for endometrial cancer, with a multivariable RR of 0.22 171
(95% CI: 0.13, 0.40) among ever users compared to never users of LNG-IUS. The final model for 172
endometrial cancer adjusted for age at start of follow-up, BMI, physical activity level, OC use, parity, and 173
menopausal status at start of follow-up. The stratified analysis showed that among ever users of OCs, 174
ever users of LNG-IUS had a RR of endometrial cancer of 0.34 (95% CI: 0.18, 0.65) compared to never 175
users of LNG-IUS. Among never users of OC, ever users of LNG-IUS had a RR of 0.08 (95% CI: 0.02, 0.34 176
compared to never users of LNG-IUS. However, these estimates were not significantly different 177
(pheterogeneity = 0.18).
178
For breast cancer, both the age-adjusted and the final adjusted model, which included age at start of 179
follow-up, BMI, physical activity level, maternal history of breast cancer, OC use, and menopausal status 180
at start of follow-up, showed no association with LNG-IUS use. The incidence rate of breast cancer was 181
275.7 per 100 000 PY among ever users of LNG-IUS and 281.6 per 100 000 PY among never users. The 182
multivariable-adjusted RR of breast cancer among ever users of LNG-IUS was 1.03 (95% CI: 0.91, 1.17).
183
Compared to never users, current users of LNG-IUS had a multivariable RR of breast cancer of 0.97 184
(95% CI: 0.80, 1.19) and former users a RR of 0.79 (95% CI: 0.64, 0.98). Stratified by duration, ever users 185
9 with up to 5 years of use had a multivariable RR of 1.06 (95% CI: 0.91, 1.24) compared to never users.
186
Those with more than 5 years of use had a RR of 0.88 (95% CI: 0.68, 1.16). Among ever users of LNG-IUS 187
with breast cancer, mean time since LNG-IUS cessation was 7.5 (SD 4.4) years (n=237). For ever users of 188
LNG-IUS not diagnosed with cancer, mean time since cessation of use was 12.5 (SD 3.3) years.
189
When all cancers were added together to produce an estimate of the total effect of LNG-IUS use, in 190
ever users the RR of any hormone-related cancer was 0.86 (95% CI: 0.77, 0.97) compared to never users.
191 192 193
10 DISCUSSION
194
In this population-based prospective cohort study, women who reported ever use of LNG-IUS showed a 195
strongly reduced risk of both ovarian and endometrial cancer compared to those who did not. LNG-IUS 196
use was not associated with an increased risk of breast cancer.
197 198
Levonorgestrel and risk of ovarian cancer 199
Several studies have investigated the association between the use of intrauterine devices and 200
ovarian cancer, but most did not include LNG-IUS users, save one American, population-based, case- 201
control study, which consisted of 104 cases and 299 controls. This study included 14 LNG-IUS users, and 202
found a negative association between ever use of intrauterine device and ovarian cancer. When 203
analyzed by duration, only 4 or fewer years of use was protective (19).A Chinese prospective cohort 204
study that may have included LNG-IUS users found no association (20).
205
Two prospective cohort studies by Soini et al. described the association between LNG-IUS use and 206
ovarian cancer (12, 13). The most recent study (12) was based on 77 ovarian cancer cases occurring in a 207
cohort of 93 843 women who had been prescribed LNG-IUS for menorrhagia. The study did not adjust 208
for risk factors, and reported that the age-adjusted SIR of ovarian cancer among women with up to 5 209
years of LNG-IUS use was 0.59 (95% CI: 0.47, 0.73). Longer duration of use did not decrease the risk 210
much further. When the entire follow-up period was taken into account, the SIR was 0.49 (95% CI: 0.24, 211
0.87) for mucinous, 0.55 (95% CI: 0.28, 0.98) for endometrioid, and 0.75 (95% CI: 0.55, 0.99) for serous 212
ovarian carcinoma. After adjusting for important risk factors, our findings confirm those of Soini et al., 213
and although our sample size did not permit analyses on histological subtypes, our adjusted results 214
strengthen the evidence of a causal association between LNG-IUS and decreased risk of ovarian cancer.
215
It is generally assumed that combined OCs prevent ovarian cancer by inhibiting ovulation (21) and 216
possibly by reducing menstrual bleeding (22). Sparse menstruations lead to less retrograde 217
11 menstruation, which, by implanting as endometriosis, is thought to be a source of either endometrioid 218
carcinoma, clear cell carcinoma, or possibly low-grade serous carcinoma (23). By other mechanisms, 219
retrograde menstruation and follicular fluid released during ovulation may induce serous tubal 220
intraepithelial carcinoma (22), which potentially could enter the ruptured ovarian epithelium and, 221
stimulated by the hormone-rich milieu of the ovary, cause high-grade serous carcinoma (24).
222
Levonorgestrel is a potent progestin. LNG-IUS used in Norway at the time questionnaires were 223
completed initially release 20 µg LNG per day, decreasing to 11 µg/day for an average of 14 µg/day over 224
a five-year period (25). LNG-IUS exerts its contraceptive effect by suppressing the endometrium, 225
thickening the cervical mucus, and, partly, by inhibiting ovulation through the hypothalamic-pituitary 226
axis (26). Most LNG-IUS users have light menstruations and 20-50% become amenorrheic (27). In the 227
present study, 30% of LNG-IUS users had unknown menopausal status, compared to 5% of non-users. In 228
an ultrasound study of 22 women, of which one-third were amenorrheic after 7 or more years of LNG- 229
IUS use, approximately 30% of amenorrheic women and 60% of still menstruating women had ovulatory 230
cycles with follicular rupture (26).
231
Risch (28) argued that, since the protective effect of progestin-only contraceptives, which do not 232
completely suppress ovulation, is comparable to the effect of combined OCs on ovarian cancer, 233
progestogens likely have a direct anti-tumorigenic effect on ovarian cancer. Such a concept was 234
supported by Merritt et al. (29) notably with regard to high natural progesterone levels during 235
pregnancy, though the effects of natural progesterone and those of synthetic progestins are not 236
superimposable. The LNG-IUS alleviates symptoms of endometriosis, and Lockhat et al. (30) showed 237
that in addition to the vascular delivery of levonorgestrel to endometriotic implants, direct contact with 238
levonorgestrel in peritoneal fluid (transferred to this fluid via blood, not by diffusion from the uterine 239
cavity) likely plays a significant role. A similar direct effect on ovarian tumors or tumor precursor cells is 240
also possible (31). This hypothesis, however, does not correspond with a Danish population-based case- 241
12 control study (21) nor with a previous study from the NOWAC cohort (32), both of which found that only 242
use of combined OCs, not oral progestogens alone, prevents ovarian cancer. Faber et al. (21) concluded 243
that OCs prevent ovarian cancer through the inhibition of ovulation. It is plausible that the preventive 244
effect of LNG-IUS on ovarian cancer works through partial inhibition of both ovulation and 245
menstruation.
246 247
Levonorgestrel and risk of endometrial cancer 248
Our adjusted results also confirm the observations of Soini et al. (13) for endometrial cancer. That 249
study adjusted for smoking, diet and alcohol consumption, socioeconomic status, and physical activity, 250
and reported a SIR of endometrial cancer of 0.46 (95% CI: 0.33, 0.64) in LNG-IUS users compared to the 251
general population. In a pooled analysis of four cohort and 14 case-control studies, Felix et al. (33) 252
calculated the association between use of different intrauterine devices and the risk of endometrial 253
cancer and found no association with LNG-IUS. However, due to the low number of women in the LNG- 254
IUS exposure group, they disregarded this result and called for further studies.
255 256
The anti-proliferative effect of LNG-IUS is superior to that of oral progestins in the treatment of 257
endometrial hyperplasia (15), and a protective effect of this device on endometrial cancer in the general 258
population is to be expected. Our results indicate the size of the risk reduction in a cohort 259
representative of the general population. Since the proportion of ever users of OCs was significantly 260
different among ever and never users of LNG-IUS, we performed an analysis stratified by ever OC use.
261
The difference was non-significant, but suggestive of a stronger protective effect of LNG-IUS among 262
never users of OCs.
263 264
13 Levonorgestrel and risk of breast cancer
265
Contrary to Soini et al. (14), we did not observe an increased risk of breast cancer among LNG-IUS 266
users. Soini et al. (14) reported a clear increased risk of certain types of breast cancer, but did not 267
present SIRs of total breast cancer. In the earlier study by Soini et al. (13), the SIR of total breast cancer 268
was 1.19 (95% CI: 1.13, 1.25). In all three studies by Soini et al. (12-14) follow-up ended at age 55 years.
269
The discrepancy between our findings and those of Soini et al. (14) could be due to their lack of 270
adjustment, although adjustment had little effect on our estimates.
271
In a recent nested case-control study of women in the Norwegian breast cancer screening program 272
(aged 50-69 years), Ellingjord-Dale et al. (34) did not find an association between duration of IUD use 273
and overall risk of breast cancer by duration of use (in intervals), although there was a statistically 274
significant trend. The results indicated increased and decreased risks of different breast cancer 275
subtypes. This study did not differentiate between types of intrauterine devices, but assuming a 276
population-representative sample and data collected from 2006 onwards, LNG-IUS users constituted a 277
large fraction of intrauterine device users (11). When we stratified on duration of use (up to 5 and more 278
than 5 years), we observed no association with breast cancer in either stratum. We did not study breast 279
cancer subtypes, and we did not test for trend.
280
A recent prospective cohort study showed that current and recent users of LNG-IUS had an 281
increased risk of breast cancer compared to never users of hormonal contraceptives (RR 1.21; 95% CI:
282
1.11, 1.33). This study included all women aged 15-49 in Denmark, and adjusted for age, calendar year, 283
education, polycystic ovary syndrome, endometriosis, parity, and family history of premenopausal 284
cancer of the breast or ovary. Our null finding remained when we restricted the analyses to current 285
users of LNG-IUS. However, in our study, few participants were younger than 46 years. Moreover, the 286
mean duration of LNG-IUS use was 4 years, and average time since cessation of use was 7.5 years. When 287
Mørch et al stratified by duration of use and time since cessation, women in the corresponding category 288
14 did not have increased and risk of breast cancer. Mørch et al. found that more than 5 years of use was 289
associated with increased cancer risk, and the risk lasted up to 10 years after cessation of use (9). In our 290
analysis stratified on duration we could not reproduce this finding.
291
Among previous studies, a Finnish case-control study of 9537 breast cancer cases and 21 598 292
controls adjusted for age at menarche, smoking, alcohol use, BMI, and family history of breast cancer 293
and found a positive association between ever use of LNG-IUS and breast cancer in postmenopausal 294
women (aged 51-64), while for premenopausal women no association was observed (35). The authors 295
mentioned the possible presence of selection bias, as some practitioners, at least in Finland (this is also 296
the case in Norway), have regarded the LNG-IUS as a preferable option for women with an increased risk 297
of breast cancer.
298 299
Strengths and limitations 300
Strengths of this study include its prospective design, inclusion of lifestyle information, and a population 301
based study cohort with women who were likely using the LNG-IUS for both contraceptive and medical 302
reasons. BMI and OCs were validated by test-retest in a subset of participants (Skeie 2015, Lund, 303
Dumeaux et al. 2008), and physical activity and menopausal status by measurements (Borch 304
2012,Waaseth 2008). Number of children was validated by Lund, Kumle (17). The LNG-IUS variable was 305
not validated, nor was maternal history of breast cancer and age at menarche. Compared to non- 306
responders, responders were at a disadvantage with regard to some risk factors for the cancers in this 307
study (lower age at menarche and nulliparity), but also had favorable characteristics (proportion of ever 308
OC users and maternal history of breast cancer). We included OC use as a dichotomous variable, as 309
analyzing OC use by duration did not change the estimate of the main exposure. We did not adjust for 310
time since OC use. Insufficient adjustment for this, and for use of other hormonal contraceptives, may 311
have caused residual confounding in our estimates.
312
15 The use of cancer registry data ensured near complete follow-up of cancer cases. However, due to 313
the strong protective effect of LNG-IUS, the study had a limited number of ovarian and endometrial 314
cancer cases. We were not able to calculate specific rates by subtype, nor could we analyze duration 315
effects on these cancer types.
316
The mean age at enrollment was lower among ever users of LNG-IUS than never users. The 317
gynecological practice of removing or leaving LNG-IUS in place at the time of menopause, varies;
318
nevertheless, even if left in place, its protective effect, if any, could be transitory, potentially delaying 319
the "natural" appearance of ovarian cancer. We created a Lexis diagram of the distribution of ovarian 320
cancer incidence in both ever users and never users of LNG-IUS, which showed a lower, but parallel 321
incidence rate among all LNG-IUS users aged less than 65 years, and a decreased incidence rate among 322
those aged 65 years or over, as compared to never users. However, among ever users of LNG-IUS, there 323
was one case that occurred after age 64 years, which introduces uncertainty into the estimation.
324
This is one of the few epidemiological studies that presents data specifically on LNG-IUS use, with 325
estimates generalizable to the general female population of Norway. However, we used self-reported 326
exposure data, which introduces a risk of misclassification. Considering the prescription routines, it is 327
likely that women were counselled by their physician and required to make a choice, and thus were 328
aware of which type of intrauterine device they were using. Nevertheless, we excluded women who 329
indicated using LNG-IUS before it was on the market.
330
CONCLUSION 331
This study shows that a relatively short period of LNG-IUS use is associated with an almost halved risk of 332
ovarian cancer, while the risk of breast cancer remains unchanged. Our results are in agreement with 333
existing data, and show a negative association in a cohort of women where the majority was older than 334
in previous studies. Although these findings suggest that LNG-IUS should be considered for inclusion in 335
16 the ovarian cancer prevention strategy for normal-risk women in addition to OCs (36), an updated meta- 336
analysis of the effect of LNG-IUS on breast cancer is needed before firm conclusions can be drawn.
337 338 339
Author contributions 340
EL and HMB conceived the study. MJ contributed to designing the analyses, interpreted results and 341
drafted the paper. EL and JCT oversaw the analyses, interpreted results and critically revised the paper.
342
TB designed the analyses, carried out analyses, and interpreted the results. HMB carried out preliminary 343
analyses, and MAA carried out final analyses. EL is the PI of the NOWAC Study. All authors read and 344
approved the final manuscript.
345
Acknowledgements 346
While employed at UiT, Nicolle Mode contributed to the design of, and scripts for, statistical analysis, as 347
well as contributing text to this paper. The authors are supported by the Faculty of Health Sciences, UiT 348
The Arctic University of Norway, Tromsø, Norway. JCT has a full-time position at the Medical Faculty, 349
Paris Descartes University, and is the beneficiary of a part-time position at UiT. The funding bodies had 350
no role in the design, collection, analysis, or interpretation of data; in the writing of the manuscript, or 351
the decision to submit the manuscript for publication. We sincerely thank the women who participate in 352
the NOWAC Study.
353
Conflict of interest statement 354
We declare that we have no conflicting interests.
355 356 357
17 REFERENCES
358
1. Zou G. A Modified Poisson Regression Approach to Prospective Studies with Binary Data. Am J 359 Epidemiol. 2004;159(7):702-6.
360 2. Ferlay J SI, Ervik M, Dikshit R, Eser S, Mathers C, Rebelo M, Parkin DM, Forman D, Bray, F.
361 GLOBOCAN 2012 v1.0, Cancer Incidence and Mortality Worldwide: IARC CancerBase No. 11. Lyon, 362 France: International Agency for Research on Cancer (IARC); 2013. Contract No.: 08-20.
363 3. SEER. Cancer Stat Facts: Ovarian Cancer: National Cancer Institute - The Surveillance, 364 Epidemiology, and End Results (SEER) Program 2017 [Available
365 from: https://seer.cancer.gov/statfacts/html/ovary.html.
366 4. Cancer Registry of Norway. Cancer in Norway 2015 - Cancer incidence, mortality, survival and 367 prevalence in Norway. Oslo: Cancer Registry of Norway; 2016.
368 5. Long Roche KC, Abu-Rustum NR, Nourmoussavi M, Zivanovic O. Risk-reducing salpingectomy: Let 369 us be opportunistic. Cancer. 2017;123(10):1714-20.
370 6. Beral V, Doll R, Hermon C, Peto R, Reeves G. Ovarian cancer and oral contraceptives:
371 collaborative reanalysis of data from 45 epidemiological studies including 23,257 women with ovarian 372 cancer and 87,303 controls. Lancet. 2008;371(9609):303-14.
373 7. Dossus L, Allen N, Kaaks R, Bakken K, Lund E, Tjonneland A, et al. Reproductive risk factors and 374 endometrial cancer: the European Prospective Investigation into Cancer and Nutrition. Int J Cancer.
375 2010;127(2):442-51.
376 8. Collaborative Group on Hormonal Factors in Breast Cancer. Breast cancer and hormonal
377 contraceptives: collaborative reanalysis of individual data on 53 297 women with breast cancer and 100 378 239 women without breast cancer from 54 epidemiological studies. The Lancet. 1996;347(9017):1713- 379 27.
380 9. Mørch LS, Skovlund CW, Hannaford PC, Iversen L, Fielding S, Lidegaard Ø. Contemporary 381 Hormonal Contraception and the Risk of Breast Cancer. N Engl J Med. 2017;377(23):2228-39.
382 10. Havrilesky LJ GJ, Moorman PG, Coeytaux RR, Peragallo Urrutia R, Lowery WJ, Dinan M, McBroom 383 AJ, Wing L, Musty MD, Lallinger KR, Hasselblad V, Sanders GD, Myers ER. Oral Contraceptive Use for the 384 Primary Prevention of Ovarian Cancer. Rockville, MD: Agency for Healthcare Research and Quality.: Duke 385 Evidence-based Practice Center; 2013 June. Report No.: Evidence Report/Technology Assessment No.
386 212. Contract No.: AHRQ Publication No. 13-E002-EF.
387 11. Lindh I, Skjeldestad FE, Gemzell-Danielsson K, Heikinheimo O, Hognert H, Milsom I, et al.
388 Contraceptive use in the Nordic countries. Acta Obstet Gynecol Scand. 2017;96(1):19-28.
389 12. Soini T, Hurskainen R, Grenman S, Maenpaa J, Paavonen J, Pukkala E. Impact of levonorgestrel- 390 releasing intrauterine system use on the cancer risk of the ovary and fallopian tube. Acta Oncol. 2016:1- 391 4.
392 13. Soini T, Hurskainen R, Grenman S, Maenpaa J, Paavonen J, Pukkala E. Cancer risk in women 393 using the levonorgestrel-releasing intrauterine system in Finland. Obstet Gynecol. 2014;124(2 Pt 1):292- 394 9.
395 14. Soini T, Hurskainen R, Grénman S, Mäenpää J, Paavonen J, Joensuu H, et al. Levonorgestrel- 396 releasing intrauterine system and the risk of breast cancer: A nationwide cohort study. Acta Oncol.
397 2016;55(2):188-92.
398 15. Orbo A, Vereide A, Arnes M, Pettersen I, Straume B. Levonorgestrel-impregnated intrauterine 399 device as treatment for endometrial hyperplasia: a national multicentre randomised trial. BJOG.
400 2014;121(4):477-86.
401 16. Lund E, Dumeaux V, Braaten T, Hjartaker A, Engeset D, Skeie G, et al. Cohort profile: The 402 Norwegian Women and Cancer Study--NOWAC--Kvinner og kreft. Int J Epidemiol. 2008;37(1):36-41.
403
18 17. Lund E, Kumle M, Braaten T, Hjartaker A, Bakken K, Eggen E, et al. External validity in a
404 population-based national prospective study--the Norwegian Women and Cancer Study (NOWAC).
405 Cancer Causes Control. 2003;14(10):1001-8.
406 18. Stewart B, Wild C, editors. World Cancer Report 2014. Lyon, France: International Agency for 407 Research on Cancer; 2014.
408 19. Ness RB, Dodge RC, Edwards RP, Baker JA, Moysich KB. Contraception Methods, Beyond Oral 409 Contraceptives and Tubal Ligation, and Risk of Ovarian Cancer. Ann Epidemiol. 2011;21(3):188-96.
410 20. Dorjgochoo T, Shu XO, Li HL, Qian HZ, Yang G, Cai H, et al. Use of oral contraceptives, 411 intrauterine devices and tubal sterilization and cancer risk in a large prospective study, from 1996 to 412 2006. Int J Cancer. 2009;124(10):2442-9.
413 21. Faber MT, Jensen A, Frederiksen K, Glud E, Hogdall E, Hogdall C, et al. Oral contraceptive use and 414 impact of cumulative intake of estrogen and progestin on risk of ovarian cancer. Cancer Causes Control.
415 2013;24(12):2197-206.
416 22. Vercellini P, Crosignani P, Somigliana E, Vigano P, Buggio L, Bolis G, et al. The 'incessant 417 menstruation' hypothesis: a mechanistic ovarian cancer model with implications for prevention. Hum 418 Reprod. 2011;26(9):2262-73.
419 23. Pearce CL, Templeman C, Rossing MA, Lee A, Near AM, Webb PM, et al. Association between 420 endometriosis and risk of histological subtypes of ovarian cancer: a pooled analysis of case–control 421 studies. The Lancet Oncology. 2012;13(4):385-94.
422 24. Kurman RJ, Shih Ie M. The origin and pathogenesis of epithelial ovarian cancer: a proposed 423 unifying theory. Am J Surg Pathol. 2010;34(3):433-43.
424 25. ESHRE Capri Workshop Group. Intrauterine devices and intrauterine systems. Hum Reprod 425 Update. 2008;14(3):197-208.
426 26. Barbosa I, Olsson SE, Odlind V, Goncalves T, Coutinho E. Ovarian function after seven years' use 427 of a levonorgestrel IUD. Adv Contracept. 1995;11(2):85-95.
428 27. Hidalgo M, Bahamondes L, Perrotti M, Diaz J, Dantas-Monteiro C, Petta C. Bleeding patterns and 429 clinical performance of the levonorgestrel-releasing intrauterine system (Mirena) up to two years1 430 1Mirena is a registered trademark of Leiras Oy, Turku, Finland. Contraception. 2002;65(2):129-32.
431 28. Risch HA. Hormonal Etiology of Epithelial Ovarian Cancer, With a Hypothesis Concerning the 432 Role of Androgens and Progesterone. JNCI: Journal of the National Cancer Institute. 1998;90(23):1774- 433 86.
434 29. Merritt MA, De Pari M, Vitonis AF, Titus LJ, Cramer DW, Terry KL. Reproductive characteristics in 435 relation to ovarian cancer risk by histologic pathways. Hum Reprod. 2013;28(5):1406-17.
436 30. Lockhat FB, Emembolu JE, Konje JC. Serum and peritoneal fluid levels of levonorgestrel in 437 women with endometriosis who were treated with an intrauterine contraceptive device containing 438 levonorgestrel. Fertil Steril. 2005;83(2):398-404.
439 31. Rodriguez GC, Nagarsheth NP, Lee KL, Bentley RC, Walmer DK, Cline M, et al. Progestin-induced 440 apoptosis in the Macaque ovarian epithelium: differential regulation of transforming growth factor-beta.
441 J Natl Cancer Inst. 2002;94(1):50-60.
442 32. Kumle M, Weiderpass E, Braaten T, Adami HO, Lund E. Risk for invasive and borderline epithelial 443 ovarian neoplasias following use of hormonal contraceptives: the Norwegian-Swedish Women's Lifestyle 444 and Health Cohort Study. Br J Cancer. 2004;90(7):1386-91.
445 33. Felix AS, Gaudet MM, Vecchia CL, Nagle CM, Shu XO, Weiderpass E, et al. Intrauterine devices 446 and endometrial cancer risk: A pooled analysis of the Epidemiology of Endometrial Cancer Consortium.
447 Int J Cancer. 2014.
448 34. Ellingjord-Dale M, Vos L, Tretli S, Hofvind S, dos-Santos-Silva I, Ursin G. Parity, hormones and 449 breast cancer subtypes - results from a large nested case-control study in a national screening program.
450 Breast Cancer Res. 2017;19(1):10.
451
19 35. Heikkinen S, Koskenvuo M, Malila N, Sarkeala T, Pukkala E, Pitkäniemi J. Use of exogenous 452 hormones and the risk of breast cancer: results from self-reported survey data with validity assessment.
453 Cancer Causes Control. 2016;27(2):249-58.
454 36. Walker JL, Powell CB, Chen L-m, Carter J, Bae Jump VL, Parker LP, et al. Society of Gynecologic 455 Oncology recommendations for the prevention of ovarian cancer. Cancer. 2015;121(13):2108-20.
456 457 458 459 460
Table 1 Characteristics of ever users (N=9144) and never users (N=95 174) of the levonorgestrel- releasing intrauterine system (LNG-IUS) in the Norwegian Women and Cancer Study, 1998-2015
* P-value from a chi-square test of independence, excluding missing value
Characteristics LNG-IUS
Ever users Never users p-value* Age at start of follow-up
(years) 41-45 1271 14 % 11177 12 % <0.01
46-50 3855 42 % 21581 23 %
51-55 3051 33 % 30526 32 %
56-60 795 9 % 18589 20 %
61-65 145 2 % 7811 8 %
66-70 20 <1 % 3012 3 %
71-76 7 <1 % 2478 3 %
Body mass index (kg/m2) <25 5295 58 % 54133 57 % 0.18
≥25 3637 40 % 38306 40 %
missing 212 2735
Physical activity level very low 248 3 % 3506 4 % <0.01
Low 1518 17 % 18200 19 %
intermediate 3479 38 % 36971 39 %
High 2972 33 % 23871 25 %
very high 556 6 % 4796 5 %
missing 371 7830
Maternal history of breast
cancer Yes 478 5 % 5032 5 % 0.80
Age at menarche (years) <12 811 9 % 8428 9 % 0.01
12-14 6646 73 % 67897 71 %
≥15 1543 17 % 17364 18 %
missing 144 1485
Ever use of oral contraceptives
Yes 6476 71 % 52259 55 % <0.01
Parity None 307 3 % 9231 10 % <0.01
1-2 5502 60 % 49935 52 %
3-4 3173 35 % 32762 34 %
≥5 162 2 % 3246 3 %
Menopausal status at
enrollment Pre 2125 23 % 24323 26 % <0.01
Peri 1206 13 % 8533 9 %
Post 3060 33 % 57128 60 %
Unknown 2753 30 % 5190 5 %
Table 2 Site-specific cancer incidence rates and relative risks comparing ever users (person-years [PY] =107 701) and never users (PY=1 197 734) of the levonorgestrel-releasing intrauterine system (LNG-IUS) in the Norwegian Women and Cancer Study
Cancer type LNG-IUS user
status
Cancer cases Incidence rate per 100 000 PY (95% CI)
Age-adjusted RR (95% CI)
Multivariable-adjusted RR (95% CI) Epithelial ovarian Ever
Never
18 457
16.7 (9.9, 26.4) 38.1 (34.7, 41.8)
0.49 (0.30, 0.82) 0.53 (0.32, 0.88)a
Endometrial Ever
Never
15 839
13.9 (7.8, 23.0) 70.0 (65.4, 74.9)
0.19 (0.11, 0.40) 0.22 (0.13, 0.40)b
Breast Ever
Never
297 3373
275.7 (245.3, 309.0) 281.6 (272.2, 291.3)
1.02 (0.90, 1.15) 1.03 (0.91, 1.17)c
Combined (ovarian, breast, endometrial)
Ever Never
330 4669
306.4 (274.2, 341.3) 389.7 (378.7, 401.2)
0.84 (0.74, 0.94) 0.86(0.77, 0.97)d
a Adjusted for OC use, age at start of follow-up, menopausal status at start of follow-up, parity
b Adjusted for OC use, age at start of follow-up, menopause status at start of follow-up, BMI, physical activity, parity
c Adjusted for OC use, age at start of follow-up, maternal history of breast cancer, BMI, physical activity, menopause status at start of follow-up
d Adjusted for OC use, age at start of follow-up, maternal history of breast cancer, BMI, physical activity, menopause status at start of follow-up, parity RR=relative risk; CI=confidence interval; BMI=body mass index; OC=oral contraceptive
1 Supplementary table 1 Selected characteristics of responders and non-responders to the question ‘Have you ever used a hormone intrauterine device (IUD)?’ in the Norwegian Women and Cancer Study, 1998- 2015
* P-value from a chi-square test of independence, excluding missing value
Characteristics Have you ever used a hormone IUD?
Responders 104 380
Non-responders
15 442 p-value* Maternal history of breast
cancer Yes 5515 5 % 940 6 % <.001
Age at menarche (years) <12 9246 9 % 1318 8 % <.001
12-14 74584 71 % 10767 70 %
≥15 18920 18 % 3030 20 %
missing 1630 2 % 327 2 %
Ever use of oral contraceptives
Yes 58761 56 % 8366 54 % <.001
Parity None 9547 9 % 939 6 % <.001
1-2 55466 53 % 8546 55 %
3-4 35957 35% 5535 36 %
≥5 3410 3 % 422 3 %